Principles of employing a self-organizing map as a frequent itemset miner

This work proposes a theoretical guideline in the specific area of Frequent Itemset Mining (FIM). It supports the hypothesis that the use of neural network technology for the problem of Association Rule Mining (ARM) is feasible, especially for the task of generating frequent itemsets and its variants (e.g. Maximal and closed). We define some characteristics which any neural network must have if we would want to employ it for the task of FIM. Principally, we interpret the results of experimenting with a Self-Organizing Map (SOM) for this specific data mining technique.